Agentic AI Atlas

II.

Specialization reference

specialization:cc10x

Reading · 3 min

CC10X reference

Intelligent workflow orchestration for Claude Code development. CC10X routes all development requests through a single entry point that detects intent and dispatches to specialized workflows with evidence-first validation, TDD enforcement, and persistent session memory.

Specializationwiki/library/cc10x.mdOutgoing · 1Incoming · 2

CC10X Methodology

Core Principles

1. **Single Entry Point**: The cc10x-router handles all requests -- no manual skill selection 2. **Evidence-First**: All claims backed by logs, test results, or exit codes (zero = success) 3. **TDD Enforcement**: RED-GREEN-REFACTOR is mandatory during builds; tests are never skipped 4. **Memory Persistence**: Context survives across sessions via three memory surfaces 5. **Confidence Gating**: Issues reported must reach >=80% confidence; vague feedback rejected 6. **Parallel Execution**: Agents run simultaneously when possible (~30-50% faster)

Workflows

Workflow	Trigger	Agent Chain
BUILD	build, implement, create, make, write, add	component-builder -> [code-reviewer \
DEBUG	debug, fix, error, bug, broken, troubleshoot	bug-investigator -> code-reviewer -> integration-verifier
REVIEW	review, audit, check, analyze, assess	code-reviewer (multi-dimensional)
PLAN	plan, design, architect, roadmap, strategy	planner (with github-researcher)

Process Files

File	Process ID	Description
`cc10x-router.js`	`methodologies/cc10x/cc10x-router`	Single entry point: intent detection, workflow dispatch, contract validation, memory persistence
`cc10x-build.js`	`methodologies/cc10x/cc10x-build`	BUILD workflow: TDD cycle, parallel review, integration verification
`cc10x-debug.js`	`methodologies/cc10x/cc10x-debug`	DEBUG workflow: log-first investigation, targeted fix, verification
`cc10x-review.js`	`methodologies/cc10x/cc10x-review`	REVIEW workflow: 4-dimension parallel analysis with confidence scoring
`cc10x-plan.js`	`methodologies/cc10x/cc10x-plan`	PLAN workflow: research, brainstorming, plan creation with build continuity

Agents (8)

Agent	Role	Used In
`cc10x-router`	Intent detection and workflow orchestration	All workflows
`component-builder`	TDD feature development	BUILD
`bug-investigator`	Log-first root cause analysis	DEBUG
`code-reviewer`	Multi-dimensional quality assessment	BUILD, DEBUG, REVIEW
`silent-failure-hunter`	Error handling gap detection	BUILD
`integration-verifier`	Evidence-backed E2E validation	BUILD, DEBUG
`planner`	Comprehensive planning with research	PLAN
`github-researcher`	GitHub repository and package research	PLAN, DEBUG

Skills (8)

Skill	Purpose
`session-memory`	Context persistence across message compaction
`verification-before-completion`	Evidence requirement enforcement
`test-driven-development`	RED-GREEN-REFACTOR enforcement
`code-generation`	Minimal, pattern-matching code output
`debugging-patterns`	Root cause analysis frameworks
`code-review-patterns`	Security, quality, performance assessment
`planning-patterns`	Structured planning methodology
`architecture-patterns`	System and API design guidance

Memory System

Three files in .claude/cc10x/:

**activeContext.md** -- current focus, decisions, learnings, blockers
**patterns.md** -- project conventions, gotchas, architectural decisions
**progress.md** -- completed work, remaining tasks, verification results

Router Contract

Every agent output includes a machine-readable contract:

**STATUS**: PASS, FAIL, APPROVE, REJECT, NEEDS_REMEDIATION
**BLOCKING**: whether workflow can proceed
**REQUIRES_REMEDIATION**: whether REM-FIX task is needed
**Issue counts**: critical and high severity tallies
**Memory notes**: items to persist for future sessions

Plan-to-Build Continuity

Plans created by the PLAN workflow are saved to docs/plans/ and referenced in session memory. The BUILD workflow reads these plans during the requirements clarification phase, ensuring consistency between strategy and execution.

Attribution

Adapted from CC10X by Rom Iluz.

CC10X reference